/*
* pgd3val: this is the value of init_mm.pgd[3] in a PV guest. It is optional.
* This to assist debug of modules in the guest. The kernel address
* space seems is always mapped, but modules are not necessarily
* mapped in any arbitraty guest cr3 that we pick if pgd3val is 0.
* Modules should always be addressible if we use cr3 from init_mm.
* Since pgd3val is already a pgd value, cr3->pgd[3], we just need to
* do 2 level lookups.
*
* NOTE: 4 level paging works for 32 PAE guests also because cpu runs in IA32-e
* mode.
* Returns: mfn for the given (pv guest) vaddr
*/
static unsigned long
dbg_pv_va2mfn(dbgva_t vaddr, struct domain *dp, uint64_t pgd3val)
{
l4_pgentry_t l4e, *l4t;
l3_pgentry_t l3e, *l3t;
l2_pgentry_t l2e, *l2t;
l1_pgentry_t l1e, *l1t;
unsigned long cr3 = (pgd3val ? pgd3val : dp->vcpu[0]->arch.cr3);
unsigned long mfn = cr3 >> PAGE_SHIFT;
DBGP2("vaddr:%lx domid:%d cr3:%lx pgd3:%lx\n", vaddr, dp->domain_id,
cr3, pgd3val);
if ( pgd3val == 0 )
{
l4t = mfn_to_virt(mfn);
l4e = l4t[l4_table_offset(vaddr)];
mfn = l4e_get_pfn(l4e);
DBGP2("l4t:%p l4to:%lx l4e:%lx mfn:%lx\n", l4t,
l4_table_offset(vaddr), l4e, mfn);
if ( !(l4e_get_flags(l4e) & _PAGE_PRESENT) )
{
DBGP1("l4 PAGE not present. vaddr:%lx cr3:%lx\n", vaddr, cr3);
return INVALID_MFN;
}
l3t = mfn_to_virt(mfn);
l3e = l3t[l3_table_offset(vaddr)];
mfn = l3e_get_pfn(l3e);
DBGP2("l3t:%p l3to:%lx l3e:%lx mfn:%lx\n", l3t,
l3_table_offset(vaddr), l3e, mfn);
if ( !(l3e_get_flags(l3e) & _PAGE_PRESENT) )
{
DBGP1("l3 PAGE not present. vaddr:%lx cr3:%lx\n", vaddr, cr3);
return INVALID_MFN;
}
}
l2t = mfn_to_virt(mfn);
l2e = l2t[l2_table_offset(vaddr)];
mfn = l2e_get_pfn(l2e);
DBGP2("l2t:%p l2to:%lx l2e:%lx mfn:%lx\n", l2t, l2_table_offset(vaddr),
l2e, mfn);
if ( !(l2e_get_flags(l2e) & _PAGE_PRESENT) ||
(l2e_get_flags(l2e) & _PAGE_PSE) )
{
DBGP1("l2 PAGE not present. vaddr:%lx cr3:%lx\n", vaddr, cr3);
return INVALID_MFN;
}
l1t = mfn_to_virt(mfn);
l1e = l1t[l1_table_offset(vaddr)];
mfn = l1e_get_pfn(l1e);
DBGP2("l1t:%p l1to:%lx l1e:%lx mfn:%lx\n", l1t, l1_table_offset(vaddr),
l1e, mfn);
return mfn_valid(mfn) ? mfn : INVALID_MFN;
}
static void build_page_tables(pgentry_t *top_table,
unsigned long *pt_pfn, void *start_va, void *end_va)
{
//
// *pt_pfn - unused pages already mapped by domain builder (512K+)
// carve new page tables/directories from *pt_pfn as needed
// start_va - end_va - range of frames to map
//
struct mmu_update mmu_updates[L1_PAGETABLE_ENTRIES];
int count = 0;
while (start_va + PAGE_SIZE <= end_va)
{
pgentry_t *tab = top_table;
unsigned int offset;
pgentry_t pte;
#if defined(__x86_64__)
offset = l4_table_offset((unsigned long)start_va);
pte = tab[offset];
if ((pte & _PAGE_PRESENT) == 0)
pte = new_pt_page(pt_pfn, tab, offset, L3_FRAME);
tab = pte_to_virt(pte);
#endif
offset = l3_table_offset((unsigned long)start_va);
pte = tab[offset];
if ((pte & _PAGE_PRESENT) == 0)
pte = new_pt_page(pt_pfn, tab, offset, L2_FRAME);
tab = pte_to_virt(pte);
offset = l2_table_offset((unsigned long)start_va);
pte = tab[offset];
if ((pte & _PAGE_PRESENT) == 0)
pte = new_pt_page(pt_pfn, tab, offset, L1_FRAME);
tab = pte_to_virt(pte);
offset = l1_table_offset((unsigned long)start_va);
pte = tab[offset];
if ((pte & _PAGE_PRESENT) == 0)
{
unsigned long pt_mfn = virt_to_mfn(tab);
mmu_updates[count].ptr = ((pgentry_t)pt_mfn << PAGE_SHIFT) + sizeof(pgentry_t)*offset;
mmu_updates[count].val = virt_to_mfn(start_va) << PAGE_SHIFT | L1_PROT;
count++;
}
start_va += PAGE_SIZE;
if (count == L1_PAGETABLE_ENTRIES || start_va + PAGE_SIZE > end_va)
{
int rc = HYPERVISOR_mmu_update(mmu_updates, count, 0, DOMID_SELF);
if (rc < 0)
fatal_error("build_page_tables: mmu_update failed: %d", rc);
count = 0;
}
}
}
void page_walk(unsigned long virt_address)
{
pgentry_t *tab = (pgentry_t *)start_info.pt_base, page;
unsigned long addr = virt_address;
printk("Pagetable walk from virt %lx, base %lx:\n", virt_address, start_info.pt_base);
page = tab[l4_table_offset(addr)];
tab = pte_to_virt(page);
printk(" L4 = %"PRIpte" (%p) [offset = %lx]\n", page, tab, l4_table_offset(addr));
page = tab[l3_table_offset(addr)];
tab = pte_to_virt(page);
printk(" L3 = %"PRIpte" (%p) [offset = %lx]\n", page, tab, l3_table_offset(addr));
page = tab[l2_table_offset(addr)];
tab = pte_to_virt(page);
printk(" L2 = %"PRIpte" (%p) [offset = %lx]\n", page, tab, l2_table_offset(addr));
page = tab[l1_table_offset(addr)];
printk(" L1 = %"PRIpte" [offset = %lx]\n", page, l1_table_offset(addr));
}
/*
* Checks if a pagetable frame is needed at 'level' to map a given
* address. Note, this function is specific to the initial page table
* building.
*/
static int need_pt_frame(unsigned long va, int level)
{
unsigned long hyp_virt_start = HYPERVISOR_VIRT_START;
#if defined(__x86_64__)
unsigned long hyp_virt_end = HYPERVISOR_VIRT_END;
#else
unsigned long hyp_virt_end = 0xffffffff;
#endif
/* In general frames will _not_ be needed if they were already
allocated to map the hypervisor into our VA space */
#if defined(__x86_64__)
if ( level == L3_FRAME )
{
if ( l4_table_offset(va) >=
l4_table_offset(hyp_virt_start) &&
l4_table_offset(va) <=
l4_table_offset(hyp_virt_end))
return 0;
return 1;
}
else
#endif
if ( level == L2_FRAME )
{
#if defined(__x86_64__)
if ( l4_table_offset(va) >=
l4_table_offset(hyp_virt_start) &&
l4_table_offset(va) <=
l4_table_offset(hyp_virt_end))
#endif
if ( l3_table_offset(va) >=
l3_table_offset(hyp_virt_start) &&
l3_table_offset(va) <=
l3_table_offset(hyp_virt_end))
return 0;
return 1;
}
else
/* Always need l1 frames */
if ( level == L1_FRAME )
return 1;
printk("ERROR: Unknown frame level %d, hypervisor %llx,%llx\n",
level, hyp_virt_start, hyp_virt_end);
return -1;
}
static int handle_cow(unsigned long addr) {
pgentry_t *tab = (pgentry_t *)start_info.pt_base, page;
unsigned long new_page;
int rc;
page = tab[l4_table_offset(addr)];
if (!(page & _PAGE_PRESENT))
return 0;
tab = pte_to_virt(page);
page = tab[l3_table_offset(addr)];
if (!(page & _PAGE_PRESENT))
return 0;
tab = pte_to_virt(page);
page = tab[l2_table_offset(addr)];
if (!(page & _PAGE_PRESENT))
return 0;
tab = pte_to_virt(page);
page = tab[l1_table_offset(addr)];
if (!(page & _PAGE_PRESENT))
return 0;
/* Only support CoW for the zero page. */
if (PHYS_PFN(page) != mfn_zero)
return 0;
new_page = alloc_pages(0);
memset((void*) new_page, 0, PAGE_SIZE);
rc = HYPERVISOR_update_va_mapping(addr & PAGE_MASK, __pte(virt_to_mach(new_page) | L1_PROT), UVMF_INVLPG);
if (!rc)
return 1;
printk("Map zero page to %lx failed: %d.\n", addr, rc);
return 0;
}
/*
* Make pt_pfn a new 'level' page table frame and hook it into the page
* table at offset in previous level MFN (pref_l_mfn). pt_pfn is a guest
* PFN.
*/
static void new_pt_frame(unsigned long *pt_pfn, unsigned long prev_l_mfn,
unsigned long offset, unsigned long level)
{
pgentry_t *tab = (pgentry_t *)start_info.pt_base;
unsigned long pt_page = (unsigned long)pfn_to_virt(*pt_pfn);
pgentry_t prot_e, prot_t;
mmu_update_t mmu_updates[1];
int rc;
prot_e = prot_t = 0;
DEBUG("Allocating new L%d pt frame for pfn=%lx, "
"prev_l_mfn=%lx, offset=%lx",
level, *pt_pfn, prev_l_mfn, offset);
/* We need to clear the page, otherwise we might fail to map it
as a page table page */
memset((void*) pt_page, 0, PAGE_SIZE);
switch ( level )
{
case L1_FRAME:
prot_e = L1_PROT;
prot_t = L2_PROT;
break;
case L2_FRAME:
prot_e = L2_PROT;
prot_t = L3_PROT;
break;
#if defined(__x86_64__)
case L3_FRAME:
prot_e = L3_PROT;
prot_t = L4_PROT;
break;
#endif
default:
printk("new_pt_frame() called with invalid level number %d\n", level);
do_exit();
break;
}
/* Make PFN a page table page */
#if defined(__x86_64__)
tab = pte_to_virt(tab[l4_table_offset(pt_page)]);
#endif
tab = pte_to_virt(tab[l3_table_offset(pt_page)]);
mmu_updates[0].ptr = (tab[l2_table_offset(pt_page)] & PAGE_MASK) +
sizeof(pgentry_t) * l1_table_offset(pt_page);
mmu_updates[0].val = (pgentry_t)pfn_to_mfn(*pt_pfn) << PAGE_SHIFT |
(prot_e & ~_PAGE_RW);
if ( (rc = HYPERVISOR_mmu_update(mmu_updates, 1, NULL, DOMID_SELF)) < 0 )
{
printk("ERROR: PTE for new page table page could not be updated\n");
printk(" mmu_update failed with rc=%d\n", rc);
do_exit();
}
/* Hook the new page table page into the hierarchy */
mmu_updates[0].ptr =
((pgentry_t)prev_l_mfn << PAGE_SHIFT) + sizeof(pgentry_t) * offset;
mmu_updates[0].val = (pgentry_t)pfn_to_mfn(*pt_pfn) << PAGE_SHIFT | prot_t;
if ( (rc = HYPERVISOR_mmu_update(mmu_updates, 1, NULL, DOMID_SELF)) < 0 )
{
printk("ERROR: mmu_update failed with rc=%d\n", rc);
do_exit();
}
*pt_pfn += 1;
}
static mfn_t
p2m_gfn_to_mfn(struct p2m_domain *p2m, unsigned long gfn,
p2m_type_t *t, p2m_access_t *a, p2m_query_t q,
unsigned int *page_order)
{
mfn_t mfn;
paddr_t addr = ((paddr_t)gfn) << PAGE_SHIFT;
l2_pgentry_t *l2e;
l1_pgentry_t *l1e;
unsigned long l1e_flags;
p2m_type_t l1t;
ASSERT(paging_mode_translate(p2m->domain));
/* XXX This is for compatibility with the old model, where anything not
* XXX marked as RAM was considered to be emulated MMIO space.
* XXX Once we start explicitly registering MMIO regions in the p2m
* XXX we will return p2m_invalid for unmapped gfns */
*t = p2m_mmio_dm;
/* Not implemented except with EPT */
*a = p2m_access_rwx;
if ( gfn > p2m->max_mapped_pfn )
/* This pfn is higher than the highest the p2m map currently holds */
return _mfn(INVALID_MFN);
/* Use the fast path with the linear mapping if we can */
if ( p2m == p2m_get_hostp2m(current->domain) )
return p2m_gfn_to_mfn_current(p2m, gfn, t, a, q, page_order);
mfn = pagetable_get_mfn(p2m_get_pagetable(p2m));
#if CONFIG_PAGING_LEVELS >= 4
{
l4_pgentry_t *l4e = map_domain_page(mfn_x(mfn));
l4e += l4_table_offset(addr);
if ( (l4e_get_flags(*l4e) & _PAGE_PRESENT) == 0 )
{
unmap_domain_page(l4e);
return _mfn(INVALID_MFN);
}
mfn = _mfn(l4e_get_pfn(*l4e));
unmap_domain_page(l4e);
}
#endif
{
l3_pgentry_t *l3e = map_domain_page(mfn_x(mfn));
#if CONFIG_PAGING_LEVELS == 3
/* On PAE hosts the p2m has eight l3 entries, not four (see
* shadow_set_p2m_entry()) so we can't use l3_table_offset.
* Instead, just count the number of l3es from zero. It's safe
* to do this because we already checked that the gfn is within
* the bounds of the p2m. */
l3e += (addr >> L3_PAGETABLE_SHIFT);
#else
l3e += l3_table_offset(addr);
#endif
pod_retry_l3:
if ( (l3e_get_flags(*l3e) & _PAGE_PRESENT) == 0 )
{
if ( p2m_flags_to_type(l3e_get_flags(*l3e)) == p2m_populate_on_demand )
{
if ( q & P2M_ALLOC )
{
if ( !p2m_pod_demand_populate(p2m, gfn, PAGE_ORDER_1G, q) )
goto pod_retry_l3;
gdprintk(XENLOG_ERR, "%s: Allocate 1GB failed!\n", __func__);
}
else
*t = p2m_populate_on_demand;
}
unmap_domain_page(l3e);
return _mfn(INVALID_MFN);
}
else if ( (l3e_get_flags(*l3e) & _PAGE_PSE) )
{
mfn = _mfn(l3e_get_pfn(*l3e) +
l2_table_offset(addr) * L1_PAGETABLE_ENTRIES +
l1_table_offset(addr));
*t = p2m_flags_to_type(l3e_get_flags(*l3e));
unmap_domain_page(l3e);
ASSERT(mfn_valid(mfn) || !p2m_is_ram(*t));
if ( page_order )
*page_order = PAGE_ORDER_1G;
return (p2m_is_valid(*t)) ? mfn : _mfn(INVALID_MFN);
}
mfn = _mfn(l3e_get_pfn(*l3e));
unmap_domain_page(l3e);
}
l2e = map_domain_page(mfn_x(mfn));
l2e += l2_table_offset(addr);
pod_retry_l2:
if ( (l2e_get_flags(*l2e) & _PAGE_PRESENT) == 0 )
{
/* PoD: Try to populate a 2-meg chunk */
if ( p2m_flags_to_type(l2e_get_flags(*l2e)) == p2m_populate_on_demand )
{
if ( q & P2M_ALLOC ) {
if ( !p2m_pod_demand_populate(p2m, gfn, PAGE_ORDER_2M, q) )
goto pod_retry_l2;
} else
*t = p2m_populate_on_demand;
}
unmap_domain_page(l2e);
return _mfn(INVALID_MFN);
}
else if ( (l2e_get_flags(*l2e) & _PAGE_PSE) )
{
mfn = _mfn(l2e_get_pfn(*l2e) + l1_table_offset(addr));
*t = p2m_flags_to_type(l2e_get_flags(*l2e));
unmap_domain_page(l2e);
ASSERT(mfn_valid(mfn) || !p2m_is_ram(*t));
if ( page_order )
*page_order = PAGE_ORDER_2M;
return (p2m_is_valid(*t)) ? mfn : _mfn(INVALID_MFN);
}
mfn = _mfn(l2e_get_pfn(*l2e));
unmap_domain_page(l2e);
l1e = map_domain_page(mfn_x(mfn));
l1e += l1_table_offset(addr);
pod_retry_l1:
l1e_flags = l1e_get_flags(*l1e);
l1t = p2m_flags_to_type(l1e_flags);
if ( ((l1e_flags & _PAGE_PRESENT) == 0) && (!p2m_is_paging(l1t)) )
{
/* PoD: Try to populate */
if ( l1t == p2m_populate_on_demand )
{
if ( q & P2M_ALLOC ) {
if ( !p2m_pod_demand_populate(p2m, gfn, PAGE_ORDER_4K, q) )
goto pod_retry_l1;
} else
*t = p2m_populate_on_demand;
}
unmap_domain_page(l1e);
return _mfn(INVALID_MFN);
}
mfn = _mfn(l1e_get_pfn(*l1e));
*t = l1t;
unmap_domain_page(l1e);
ASSERT(mfn_valid(mfn) || !p2m_is_ram(*t) || p2m_is_paging(*t));
if ( page_order )
*page_order = PAGE_ORDER_4K;
return (p2m_is_valid(*t) || p2m_is_grant(*t)) ? mfn : _mfn(INVALID_MFN);
}
static mfn_t
p2m_pt_get_entry(struct p2m_domain *p2m, unsigned long gfn,
p2m_type_t *t, p2m_access_t *a, p2m_query_t q,
unsigned int *page_order)
{
mfn_t mfn;
paddr_t addr = ((paddr_t)gfn) << PAGE_SHIFT;
l2_pgentry_t *l2e;
l1_pgentry_t *l1e;
unsigned long l1e_flags;
p2m_type_t l1t;
ASSERT(paging_mode_translate(p2m->domain));
/* XXX This is for compatibility with the old model, where anything not
* XXX marked as RAM was considered to be emulated MMIO space.
* XXX Once we start explicitly registering MMIO regions in the p2m
* XXX we will return p2m_invalid for unmapped gfns */
*t = p2m_mmio_dm;
/* Not implemented except with EPT */
*a = p2m_access_rwx;
if ( gfn > p2m->max_mapped_pfn )
/* This pfn is higher than the highest the p2m map currently holds */
return _mfn(INVALID_MFN);
mfn = pagetable_get_mfn(p2m_get_pagetable(p2m));
{
l4_pgentry_t *l4e = map_domain_page(mfn_x(mfn));
l4e += l4_table_offset(addr);
if ( (l4e_get_flags(*l4e) & _PAGE_PRESENT) == 0 )
{
unmap_domain_page(l4e);
return _mfn(INVALID_MFN);
}
mfn = _mfn(l4e_get_pfn(*l4e));
unmap_domain_page(l4e);
}
{
l3_pgentry_t *l3e = map_domain_page(mfn_x(mfn));
l3e += l3_table_offset(addr);
pod_retry_l3:
if ( (l3e_get_flags(*l3e) & _PAGE_PRESENT) == 0 )
{
if ( p2m_flags_to_type(l3e_get_flags(*l3e)) == p2m_populate_on_demand )
{
if ( q & P2M_ALLOC )
{
if ( !p2m_pod_demand_populate(p2m, gfn, PAGE_ORDER_1G, q) )
goto pod_retry_l3;
gdprintk(XENLOG_ERR, "%s: Allocate 1GB failed!\n", __func__);
}
else
*t = p2m_populate_on_demand;
}
unmap_domain_page(l3e);
return _mfn(INVALID_MFN);
}
else if ( (l3e_get_flags(*l3e) & _PAGE_PSE) )
{
mfn = _mfn(l3e_get_pfn(*l3e) +
l2_table_offset(addr) * L1_PAGETABLE_ENTRIES +
l1_table_offset(addr));
*t = p2m_flags_to_type(l3e_get_flags(*l3e));
unmap_domain_page(l3e);
ASSERT(mfn_valid(mfn) || !p2m_is_ram(*t));
if ( page_order )
*page_order = PAGE_ORDER_1G;
return (p2m_is_valid(*t)) ? mfn : _mfn(INVALID_MFN);
}
mfn = _mfn(l3e_get_pfn(*l3e));
unmap_domain_page(l3e);
}
l2e = map_domain_page(mfn_x(mfn));
l2e += l2_table_offset(addr);
pod_retry_l2:
if ( (l2e_get_flags(*l2e) & _PAGE_PRESENT) == 0 )
{
/* PoD: Try to populate a 2-meg chunk */
if ( p2m_flags_to_type(l2e_get_flags(*l2e)) == p2m_populate_on_demand )
{
if ( q & P2M_ALLOC ) {
if ( !p2m_pod_demand_populate(p2m, gfn, PAGE_ORDER_2M, q) )
goto pod_retry_l2;
} else
*t = p2m_populate_on_demand;
}
unmap_domain_page(l2e);
return _mfn(INVALID_MFN);
}
else if ( (l2e_get_flags(*l2e) & _PAGE_PSE) )
{
mfn = _mfn(l2e_get_pfn(*l2e) + l1_table_offset(addr));
*t = p2m_flags_to_type(l2e_get_flags(*l2e));
unmap_domain_page(l2e);
ASSERT(mfn_valid(mfn) || !p2m_is_ram(*t));
if ( page_order )
*page_order = PAGE_ORDER_2M;
return (p2m_is_valid(*t)) ? mfn : _mfn(INVALID_MFN);
}
mfn = _mfn(l2e_get_pfn(*l2e));
unmap_domain_page(l2e);
l1e = map_domain_page(mfn_x(mfn));
l1e += l1_table_offset(addr);
pod_retry_l1:
l1e_flags = l1e_get_flags(*l1e);
l1t = p2m_flags_to_type(l1e_flags);
if ( ((l1e_flags & _PAGE_PRESENT) == 0) && (!p2m_is_paging(l1t)) )
{
/* PoD: Try to populate */
if ( l1t == p2m_populate_on_demand )
{
if ( q & P2M_ALLOC ) {
if ( !p2m_pod_demand_populate(p2m, gfn, PAGE_ORDER_4K, q) )
goto pod_retry_l1;
} else
*t = p2m_populate_on_demand;
}
unmap_domain_page(l1e);
return _mfn(INVALID_MFN);
}
mfn = _mfn(l1e_get_pfn(*l1e));
*t = l1t;
unmap_domain_page(l1e);
ASSERT(mfn_valid(mfn) || !p2m_is_ram(*t) || p2m_is_paging(*t));
if ( page_order )
*page_order = PAGE_ORDER_4K;
return (p2m_is_valid(*t) || p2m_is_grant(*t)) ? mfn : _mfn(INVALID_MFN);
}
